本研究主要目標為透過最大限度地利用現有氣象資源，為台灣開發一個可供作業化的風暴潮系集預報系統。該模型是根據數值風暴潮模型 COMCOT-SURGE、潮汐模型 TPXO 以及根據區域系集天氣預報系統 (WEPS) 和颱風預報歷史誤差統計產生集合成員的技術所驅動的，根據不同成員數量的系集組成方式，得到可以平衡預測表現與計算效能的建議。 研究中首先使用颱風最佳路徑參數資料和區域颱風天氣預報系統TWRF之分析場進行後報，以檢查在氣象場輸入準確時暴潮模式的預測表現，研究中顯示理想風場和TWRF模式分析場在不同情況下都有優勢。在系集組成方式的比較中，對2016年四次颱風事件的所有預報對生成集合成員的所有可用技術進行評估，並展示颱風登陸/最接近台灣前24小時的系集預報產品。在各個案例比較結果中表明，WEPS相關的系集組合始終低估了觀測到的水位，而路徑誤差系集組合則普遍高估了極端情況。在評估績效的指標中顯示，增加集合成員可以提高可能性的覆蓋範圍，但計算收益遞減，並且來自歷史預測誤差分佈的成員組合可以有效地執行並具有足夠的完整性。 總體而言，本研究展示了為台灣研發作業用風暴潮系集預報系統的有效方法。研究中針對模式開發提供了準確性評估、不確定性的表示方式、計算效能和預報產品解釋。該系統在幫助備災方面具有廣闊的潛力。;The main objective of this study is to develop an operational storm surge ensemble forecasting system for Taiwan by maximizing the use of existing meteorological resources. The model is driven by the numerical storm surge model COMCOT-SURGE, the tidal model TPXO, and ensemble member generation techniques based on the regional ensemble weather forecasting system (WEPS) and historical typhoon forecast error statistics. Based on different ensemble configurations with varying member counts, recommendations that balance forecast performance and computational efficiency are obtained. The study first conducts hindcasts using best track parameters and analysis fields from the regional Typhoon Weather Research and Forecast System (TWRF) to evaluate the predictive performance of the surge model given accurate meteorological inputs. The results show that the idealized wind and TWRF analysis fields have advantages in different situations. In comparing ensemble configuration approaches, all available techniques for generating ensemble members are assessed for four typhoon cases in 2016, and ensemble forecast products 24 hours before typhoon landfall/closest approach to Taiwan are displayed. The inter-case comparison indicate that the WEPS-related ensemble consistently underestimates the observed water levels, while the track error ensemble tends to overestimate. Performance metrics show that increasing ensemble members can improve possibility coverage but with diminishing returns in computation, and combinations drawing from historical forecast error distributions can perform effectively with sufficient integrity. This study demonstrates practical approaches for developing an operational storm surge ensemble forecasting system for Taiwan. The research assesses accuracy, representations of uncertainty, computational efficiency, and interpretations of forecast products about system development. Such a system holds great potential in aiding disaster preparation.
